Study On Key Techniques Of Vision Measurement For Complex Optical Surface

Visual measurement comprehensively utilizes electronics,photoconductive detection,image processing and computer technology,Machine vision was introduced to industrial measurement which could realize a rapid measurement of the object(products or parts)in terms of their three-dimensional size or location.It will boast a broad application prospect in many fields of industrial production with its prominent advantages of non-contact,high speed,high precision,strong performance of real-time and flexibility.In recent years,with constant improvement of the processing method,process control system and productive level,the object structure visual detection techniques applied to has become more complex,environmental condition is harsher and the detection needs more diversity,which will make some new challenges occur in the development of visual detection technology.This paper aims at shadows,blind areas,invalid data,under samplings,overexposure and reconstruction failure occurring on the structure surface with complicated optical characteristics during the process of visual measurement.It also puts forward some solutions of abrupt curvature change,high light ratio and high reflective surface in terms of the steps and holes.Some key technologies are mainly studied including light stripe center extraction under high reflective conditions,high dynamic range acquisition and data fusion,combining with the productive process requirements of hard-disk baseplate,and a set of 3D binocular structured light measurement system is developed based on polarized multi-exposure image fusion.The main contents include:It has analyzed the polarization properties of the scattering light on metal surface,established the polarized reflectance model of the metal surface,the relationship between polarized reflectance ratio on substrate surface of the hard disk and the polarization angle,as well as the gray value distribution of the light stripe under different polarization angles to determine the optimal polarization angle.Method of light stripe center extraction with fractional differential was put forward and the fractional differential method can be used to enhance high frequency noise particles while retain the texture features of the low frequency continuous stripe and highlight the granular characteristics of noise in terms of the light stripe imaging in the high reflective areas.The experiment demonstrated that the fractional differential could better highlight the changes of the stripe edge.According to the characteristics of strong black,strong white and local high reflection of the surface,it has proposed the method of double exposure-fusion HDR based on local energy gradient operator,where the high reflection areas and black dark areas were taken into account.Using high and low exposures to make a high dynamic range imaging could solve the problem of imaging difference caused by insufficient exposure or over exposure during the imaging process of high light ratio surface.The phenomenon of data loss and cavitation would occur when measure some surfaces with abrupt change of curvature,such as the steps and holes.Thus the dual tree complex wavelet transform(DWT)image fusion method was raised to respectively define the activity measure and match measure of the high frequency and low frequency region and calculate the fusion factor of each frequency band.Through the complementary and fusion of camera light stripe,surface curvature change can be avoided caused by data loss.A three dimensional vision measurement system for hard-disk baseplate was developed,which has been applied to the industry of hard disk baseplate.The measurement error of the system is within 8 ?m,the measurement time of single harddisk baseplate is 6s,which meets the requirement of the production on line detection of hard-disk baseplate.